Shaoting Zhu (朱少廷)

Hey, I'm Shaoting, a second-year PhD student at Institute for Interdisciplinary Information Sciences (IIIS), Tsinghua University, majoring in Computer Science. I am very fortunate to be advised by Prof. Hang Zhao in MARS Lab. Before that, I received my bachelor's degree from Zhejiang University. I was advised by Prof. Yong Liu in APRIL Lab from CSE.

I like to make handicrafts and listen to music in my spare time. Also, I love doing outdoor activities, especially traveling and road trip.

Email  /  Scholar  /  Twitter  /  Github

Recent News

Research

My research interest includes Robotics, Computer Vision and Artificial Intelligence. (*indicates equal contribution)

We propose a real-to-sim-to-real framework that leverages rapid test-time training (TTT) on novel terrains, significantly enhancing the robot's capability to traverse extremely difficult geometries.

We present a scalable, end-to-end perceptive parkour framework designed for robust humanoid hiking. Our policy enables robust traversal of complex terrains at speeds up to 2.5 m/s.

We present a framework where exteroceptive sensing is integrated into whole-body motion tracking, permitting a humanoid to perform highly dynamic, non-locomotion tasks on uneven terrain.

VR-Robo introduces a digital twin framework using 3D Gaussian Splatting for photorealistic simulation, enabling RGB-based sim-to-real transfer for robot navigation and locomotion.

MoELoco introduces a multitask locomotion framework that employs a mixture-of-experts strategy to enhance reinforcement learning across diverse tasks while leveraging compositionality to generate new skills.

RoboEngine is the first plug-and-play visual robot data augmentation toolkit. Users can effortlessly generate physics-aware robot scenes with few lines code. This enable training only in one scenes and visual generalizing to almost arbitrary scenes.

We propose a proprioception-only, two-stage training framework with goal command and a dedicated tiny trap benchmark, enabling quadruped robots to robustly traverse small obstacles.

SARO is an innovative system composed of a high-level reasoning module, a closed-loop sub-task execution module, and a low-level control policy. It enables the robot to navigate across 3D terrains and reach the goal position.

We focus on Test-Time Training (TTT) for transductive models and develop our pipeline following the SOTA methods, which consists of three steps: Base Model Training, TTT and Active Inference.

UniFace++ combines the advantages of each, ie, stability of reconstruction training from reenactment, simplicity and effectiveness of the target-oriented processing from swapping, and redefining both as target-oriented reconstruction tasks.

Given a textured face as the source and the rendered face projected from the desired 3DMM coefficients as the target, our proposed Texture-Geometry-aware Diffusion Model decomposes the complex transfer problem into multi-conditional denoising process.

Miscellanea

Template borrowed from jonbarron.